I have an audiobook, but I wasn’t satisfied with the chapters. They were just numbers, and didn’t match with the chapters in the ebook.

This was the result.

First, use ffprobe to confirm the m4a file has chapters. We don’t want to figure out the timestamps ourselves.

ffprobe Morgan_Housel_-_The_Psychology_of_Money.m4a 
ffprobe version n6.0 Copyright (c) 2007-2023 the FFmpeg developers
  built with gcc 13.2.1 (GCC) 20230801
  ...
[mov,mp4,m4a,3gp,3g2,mj2 @ 0x55cd72a08740] stream 0, timescale not set
Input #0, mov,mp4,m4a,3gp,3g2,mj2, from 'Morgan_Housel_-_The_Psychology_of_Money.m4a':
  Metadata:
    major_brand     : isom
    minor_version   : 512
    compatible_brands: isomiso2mp41
    title           : The Psychology of Money: Timeless Lessons on Wealth, Greed, and Happiness
    artist          : Morgan Housel
    composer        : Chris Hill
    album           : The Psychology of Money: Timeless Lessons on Wealth, Greed, and Happiness
    date            : 2020
    encoder         : inAudible 1.97
    comment         : Money - investing, personal finance, and business decisions - is typically taught as a math-based field, where data and formulas tell us exactly what to do. But in the real world people don’t make financial decisions on a spreadsheet. They make them at 
    genre           : Audiobook
    description     : Money - investing, personal finance, and business decisions - is typically taught as a math-based field, where data and formulas tell us exactly what to do. But in the real world people don’t make financial decisions on a spreadsheet. They make them at 
  Duration: 05:48:30.64, start: 0.000000, bitrate: 63 kb/s
  Chapters:
    Chapter #0:0: start 0.000000, end 13.606000
      Metadata:
        title           : 001
    Chapter #0:1: start 13.606000, end 804.571000
      Metadata:
        title           : 002
    Chapter #0:2: start 804.571000, end 1981.405000
      Metadata:
        title           : 003
    Chapter #0:3: start 1981.405000, end 3212.387000
      Metadata:
        title           : 004
    Chapter #0:4: start 3212.387000, end 4043.987000
      Metadata:
...

Pretty good, it contains enough metadata that I can probably rename it straight into a m4b.

# Dump the chapters, which will be in .ttxt format
$ MP4Box -dump-chap book.m4b

# Add the improved chapters back to the m4b
# MP4Box will edit the original file by default. We want that, because otherwise we will have to explicitly copy over the cover image, the artist/genre/encoder/composer metadata...
$ MP4Box -chap book.ttxt Psychology\ of\ Money\ -\ Morgan\ Housel.m4b

Perfect. The finished audiobook is a joy to browse through.

Recently I tried to use GPT-4 to analyze dove-foundation/dove in preparation for a huge refactor. This is similar to feeding in PDFs and asking questions of it.

llama-index and langchain were famous in this space (loading documents and indexing), but some people said they were overly complex, so I spent some time on griptape, simpleaichat.
https://www.youtube.com/watch?v=eQHhB1Z3fIA

But they seemed too simplistic. As I delved deeper into langchain, I didn’t find it was too complicated. Rather, it was very flexible and documentation was splendid.

I followed this tutorial.

When I asked GPT4 about the code, sometimes it would reply that not enough context was provided or that no code was provided. That’s weird, because I had gotten langchain to index the data. Time to dig deeper into what was really going on.

Code Splitting vs Text Splitting.

Turns out, when you feed text into langchain, it splits it up into chunks. And by default it splits them up by periods. Which makes sense for normal text, but not code. So you need a splitter that can understand code and split things up without mangling the meaning.

The model started to give useful output more often. Being able to see the chunks on Deeplake’s web UI helped a lot.

Must I use Deeplake?

As it turns out, the AI doesn’t have enough working memory (‘context’ is the technical word) to keep the whole codebase in mind. So people insert a database/index in between, which reads your data, determines what chunks of code are relevant to the question you’re asking and sends them together to the AI.

The tutorial told me to use Deeplake for the database/index, but I knew Deeplake was just a cloud play for something that doesn’t really need to be on the cloud (isn’t this familiar, fellow developers). Having a database/index isn’t something you need a cloud account for, you should be able to run that on your own computer.

Chroma is langchain’s built in database/index. It was terrible. I asked a question about evm.go and it didn’t even give the file to GPT4, only 2 results were relevant.

Then I tried an Ensemble retriever, BM25+FAISS. I asked the question “what does Observer.SaveAndPush()` do and paused my script to see what supporting code snippets the database/index would send to GPT4 to help make sense of my question.

pretty_print_docs(retriever.get_relevant_documents("what does Observer.SaveAndPush do?"))
Document 1 ({'source': '/home/shinichi/source/work/dove/bird/observer/observer.go'}):

func (d *Observer) SaveAndPush(event ev.EventObserved) {

    // Directly provide Job status data to the executor service
    // the "standard" way to do this would be to have the executor check in
    // with a contract on DoveChain that pulls data from EventStorage.sol
    // Specifically, the ExecutionQueue should use UELs to get Status data from
    // EventStorage.sol? Future ToDo
    if event.ChainName == "DoveGeth" &&
        event.ContractAddress == d.config.ExecutionQueueAddress.String() {

        d.modLog.Info("New ExecutionQueueEvent", "event", event)

        abi, err := abi.JSON(strings.NewReader(execution_queue.ExecutionQueueABI))
        if err != nil {
            // This shall not happen
            d.modLog.Info("OBS/EXC: ExecutionQueueAbi parse failed", "err", err)
            return
        }
----------------------------------------------------------------------------------------------------
Document 2 ({'source': '/home/shinichi/source/work/dove/bird/observer/observer.go'}):

package observer

import (
    "bytes"
    //"encoding/json"
    "strings"
    "time"

    "github.com/dove-foundation/dove/bird/config"
    "github.com/dove-foundation/dove/bird/contracts/execution_queue"
    "github.com/dove-foundation/dove/bird/contracts/handle"
    "github.com/dove-foundation/dove/bird/executor"

    "github.com/dove-foundation/dove/bird/chain_service"
    drivers "github.com/dove-foundation/dove/bird/drivers"
    eventbrowserdb "github.com/dove-foundation/dove/bird/event_api"
    pool "github.com/dove-foundation/dove/bird/eventpool"
    ev "github.com/dove-foundation/dove/bird/events"
    p2p "github.com/dove-foundation/dove/bird/p2p"
    reshare_monitor "github.com/dove-foundation/dove/bird/reshare_monitor"
    tbls "github.com/dove-foundation/dove/bird/tbls"

    logger "github.com/dove-foundation/dove/bird/logger"
    log "github.com/inconshreveable/log15"
----------------------------------------------------------------------------------------------------
Document 3 ({'source': '/home/shinichi/source/work/dove/bird/observer/observer.go'}):

func (d *Observer) reportMaliciousExecutor(executor common.Address, eventKey common.Hash) {
    self := d.config.MyAddress == executor
    if !self {
        tx, err := d.chainService.ReportBird(executor, chain_service.JOB_RELAY_FAULTY, eventKey)
        if err != nil {
            d.modLog.Error("OBS/EXEC: Failed to report malicious executor", "error", err)
        } else {
            d.modLog.Info("OBS/EXEC: Report malicious executor success", "txHash", tx)
        }
    }
}

// Verify the transaction hash updated by executor by checking whether
// executor called the correct contract and passed the correct parameters
// based on dispatched information from ExecutionQueue.sol
func (d *Observer) verifyJobTX_Save_Push(
    jobId [32]byte,
    txHash string,
    event ev.EventObserved) {

    d.modLog.Info("OBS/EXEC: VerifyJobTransaction",
        "JobId", common.BytesToHash(jobId[:]).Hex()[2:9])

    abi, err := abi.JSON(strings.NewReader(handle.HandleABI))
    if err != nil {
        d.modLog.Error("OBS/EXEC: Failed to parse HandleABI", "error", err)
        return
    }
----------------------------------------------------------------------------------------------------
Document 4 ({'source': '/home/shinichi/source/work/dove/bird/observer/observer.go'}):

var Exit = make(chan bool)

type Observer struct {
    chainConfigs    []config.ChainConfig
    config          config.MasterConfig
    watching        *ev.Watching
    eventPool       *pool.EventPool
    eventMiner      *pool.EventMiner
    transport       p2p.P2P
    modLog          log.Logger
    chainService    chain_service.ChainService
    executor        *executor.Executor
    executorAddress common.Address
    chainHandlers   map[int64]drivers.ChainHandler
    eventChannel    chan ev.EventObserved
    errorChannel    chan error
}
----------------------------------------------------------------------------------------------------
Document 5 ({'source': '/home/shinichi/source/work/dove/bird/observer/observer.go'}):

/*
This Observer takes information for _one_ chain, and then looks for events on that chain.

Note that this code is not well written - e.g. we don't filter eventsConfig
----------------------------------------------------------------------------------------------------
Document 6 ({'source': '/home/shinichi/source/work/dove/bird/event_api/event_api_interface.go'}):

package eventbrowserdb

import (
    "github.com/dove-foundation/dove/bird/events"
    "github.com/ethereum/go-ethereum/common"
)

type EventApiInterface interface {
    SaveNewEvent(ev events.EventObserved,
        coreHash common.Hash)
    SaveSubmitterEventHash(eventKey string,
        txHash string)
    SaveSubmitterError(eventKey string,
        errMsg string)
    SaveAction(observerId common.Address,
        eventKey string,
        coreHash string,
        timestamp uint64,
        actionName string)
    SaveActionWithData(observerId common.Address,
        eventKey string,
        coreHash string,
        timestamp uint64,
        actionName string,
        actionData []byte)
}
----------------------------------------------------------------------------------------------------
Document 7 ({'source': '/home/shinichi/source/work/dove/bird/cmd/root.go'}):

go ob.Start()
        <-observer.Exit // no message is ever sent to this channel, so it just serves as a block for the goroutines to run

        return nil
    },
}
----------------------------------------------------------------------------------------------------
Document 8 ({'source': '/home/shinichi/source/work/dove/bird/contracts/events_to_watch/events_to_watch.go'}):

nfirmations\",\"type\":\"uint64\"},{\"internalType\":\"uint64\",\"name\":\"ObservationStartBlock\",\"type\":\"uint64\"},{\"internalType\":\"bool\",\"name\":\"OverrideSender\",\"type\":\"bool\"}],\"internalType\":\"structEventsToWatch.Event\",\"name\":\"\",\"type\":\"tuple\"}],\"stateMutability\":\"view\",\"type\":\"function\"},{\"inputs\":[{\"internalType\":\"bytes32\",\"name\":\"_eventKey\",\"type\":\"bytes32\"}],\"name\":\"getEventKeyIndex\",\"outputs\":[{\"internalType\":\"int256\",\"name\":\"\",\"type\":\"int256\"}],\"stateMutability\":\"view\",\"type\":\"function\"},{\"inputs\":[{\"internalType\":\"bytes32\",\"name\":\"eventKey\",\"type\":\"bytes32\"}],\"name\":\"getFollowUpAction\",\"outputs\":[{\"components\":[{\"internalType\":\"string\",\"name\":\"MethodName\",\"type\":\"string\"},{\"internalType\":\"address\",\"name\":\"ContractAddress\",\"type\":\"address\"}],\"internalType\":\"structEventsToWatch.FollowUpAction\",\"name\":\"\",\"type\":\"tuple\"}],\"stateMutability\":\"view
----------------------------------------------------------------------------------------------------
Document 9 ({'source': '/home/shinichi/source/work/dove/bird/drivers/bitcoin.go'}):

*/
----------------------------------------------------------------------------------------------------
Document 10 ({'source': '/home/shinichi/source/work/dove/bird/drivers/bitcoin_test.go'}):

Even a novice programmer could see that only Documents 1, 3, 4 are relevant to the question at hand, but even then those weren’t the full function bodies, and the subfunctions that SaveAndPush() called weren’t even included, so how could GPT4 even give a satisfactory answer?

Integrating a language server that actually understands the programming language like gopls is really needed to understand what’s relevant to the question instead of a stupid search method.

Ah, Github Copilot

Then I remembered I had a Github Copilot subscription. It just makes sense that my IDE would understand enough about the code to send the relevant code to GPT.

I immediately installed Visual Studio Code and forgot about Code OSS. When it comes to cutting edge technology, open source software tends to disappoint. It’s just the way it is, unfortunately. For interoperability and commoditization, open source software is good.

Overall, AI cannot understand an entire codebase. This means it can’t help with design/architecture, but it can help with small annoying problems. With small enough code snippets, it can read and analyze code to find problems really quickly.

Supposedly Anthropic Claude-2 is better than GPT4 at coding and way cheaper (I spent about 3-4USD on this experiment) but for some reason it’s not available outside of the US.

I’ve said this before in many private events but I guess I should finally put this into writing.

The Body-State Analogy

Cells come together to form bodies, just like humans come together to form states. Blood flows between the cells, around passing value (nutrients, oxygen). If the body doesn’t appreciate a particular body part’s function, less blood flows to that part. And so it is with money too – the more valuable an industry is perceived, the more investments and resources go into it.

Taking the Analogy Further

Organs -> Specialization: just like cells come together to form organs with specific functions in the body, people or companies specialize in certain industries.
Immune System -> Regulations: regulations can be seen as the immune system of a country, keeping foreign cells from entering the body or causing harm. We all say it stifles innovation – that’s right, regulations protect the status quo.
Circulatory System -> Infrastructure: transportation networks and financial institutions are like blood vessels. Communication infrastructure would then be the nervous system.
Homeostasis -> Economic Balance: Homeostasis is the maintenance of stable internal conditions (medical). Similarly, a state also needs to maintain economic balance by managing inflation, unemployment and fiscal policies.

Crypto is a blood looking for a body

In 2019, I had a choice to make. Go into AI, or crypto? I chose crypto – because it had the potential to be more disruptive – it could potentially upheave entire countries and change the way we transact.

Fast forward to today, and bodies have successfully (as in, the country continues to exist and hasn’t been totally disrupted) managed their relationship with crypto by either regulating it away or embracing it wholeheartedly – strategies differ.

DAOs (analogous to companies) exist based on coins, but still rely on traditional infrastructure provided by nation-states to exist. There is no state-level entity that uses crypto as its blood.

Could Balaji Srinivasan’s Network State be such an entity?

Bitcoin has created a public utility (payment rails/value storage) out of thin air just by taking advantage of people’s greed. To see how far we can replicate Bitcoin’s success in other real life businesses, let’s try to decentralize several businesses.

Manufacturing a Table

The steps in making a table are: make a design, source material, work the material into shapes, assemble pieces.

Design: extremely subjective. Many people will come up with different designs, and any decision has to be made by a human with ‘good taste’.

Material sourcing: the choice of a material over another depends not just on its performance, suitability to the design, but also its price and availability. Which means a human will make this decision.

Working the material: perhaps a machine can shape the material and inspect the finished product. Realistically though a human is involved.

Assembling parts: a machine could do this, but humans are still needed for QC.

Conclusion: table manufacturing has no business being ‘decentralized’ a la Bitcoin, because the work done is physical and human taste is needed to judge almost all parts of it.

Plus, being ‘decentralized’ wouldn’t bring any benefits for the end user.

Email Hosting

End user benefit: an email service that doesn’t read its users’ email, but yet doesn’t need a user to pay money.

Steps: Setup hardware, software, hosting service: same as when setting up your own blockchain node

Hey, this already exists and is called self-hosting! But self-hosting is just running the email server for yourself, let’s go further than that and offer the service to others.

To prevent abuse we have to ask for a cost again, which could either be

• time (run the email server on your own machine)
• token (fiat money or cryptocurrency), which begs the question: how should the user earn the token. Perhaps he could host the software too, but then that would be back to self-hosting.

Conclusion: email is already decentralized, and we are simply being spoilt by Gmail being free. Everything has a cost.

Search/Recommendation Engine

End user benefit: a search/recommendation engine that isn’t corrupted by advertising, yet still manages to exist without asking users for money. A true common good.

Scraping webpages: This seems like a great task to hand off to random distributed computers, except it isn’t: the content of a URL should be objective, but isn’t – pages change with time, and even by IP geolocation. Thus the truth is subjective. We must guard against an attacker who wants to game the recommendation algorithm by submitting a fake version of his webpage. We can do this economically by requiring that tokens be staked, but it should not be the only safeguard as it depends on the token’s price.

Alright, scraping webpages takes resources and introduces subjectivity. Let’s instead record which users liked which items, which means we don’t need to know anything about the URL except that it isn’t a 404.

A like/dislike button in the web browser influences what is shown in the browser sidebar, called ‘Similar Pages’.

The user earns tokens by expressing his like/dislike for pages. To make this data available on a distributed database, he stakes tokens on his preferences, and receives the tokens as his reward for adding his preferences to the dataset.

However, what a user likes/dislikes is subjective, and any participant trying to game the system (to earn tokens, to establish false relationships between content) can hide behind that argument and the system cannot slash him for acting dishonestly. Again we find that the difficulty of verifying something external in an automated way is the problem.

OK let’s go back to scraping webpages and labeling them using an AI. It’s less subjective and therefore less complex. Companies have done this before. Maybe we can keep the centralized company but reimagine the economics.

As it turns out, there are search engines like Neeva and Kagi. Both are paid and seem to deliver less SEO clickbaity results. Perhaps we’re all just spoiled and have to pay after all.

It turns out you do need humans in the loop to make good search results after all, so at some point you will have to pay humans. According to Wikimedia Foundation Financials humans are the biggest expense item, so even if you decentralize the system enough so that it runs entirely on users’ computers, you have to pay humans to maintain it.

Humans currently maintain Bitcoin, Ethereum and Solana, so this isn’t a showstopper. So why is humanity willing to pay for decentralized computation when it’s not willing to pay for search, which builds upon computation and storage?

Because search is already available elsewhere for free, and it’s good enough, whereas immutable, censorship-resistant computation always costs, no matter where you look. Google is bad, but I can work around it, and there are other free search engines.

So just because something is immensely valuable (like water) people aren’t going to pay for it. People will only pay when their relative circumstances force them to.

Learnings

Decentralization adds massive complexity because of the need for objective, easily verifiable truth.

Bitcoin is simple because it relies on simple objective truth (mathematics) which does not rely on external data. This allows it to know for sure when someone deserves its newly minted currency.

Aligned values and being decentralized are two different things. Search engines like Neeva and Kagi show that if users pay, values are aligned.

To align values between a user and a provider, a cost must be paid, either in a token, or time (participating in running the system). If there is a token, the question is how does the user earn it, besides buying it with fiat money.

If humans are doing work in the system, they will be the bulk of the costs. Thus you will eventually need to pay money for the token (otherwise it will go down in value), or just pay money for the service like in real life.

People don’t pay for absolute value. They pay when they can’t get that value anywhere else (or not without a steep cost in time/energy).

My EDC includes an iPhone 6S, Xiaomi Redmi Note 9 Pro, and when I really want to take pictures, a Sony RX100 IV.

In a fit to optimize everything in my life, I tried to optimize this away into one device, but couldn’t – because the thought process that had landed me in this situation was very sound and still applies even today. Let me explain.

What my Android phones were like

My first smartphone was an LG Optimus One, then a HTC One S. Both companies are terrible at software updates – the One S had its last Android update only 6 months after it was released – and I had bought it at full price, new! I was so pissed I vowed never to buy HTC again. When they went under, I was glad.

Ahem. Instead of ranting about it, let me just summarize my experiences with each phone.

LG Optimus One: eye opening introduction to smartphones; SLOW; obviously bad rear camera; SLOW. The lack of software updates was not an issue because the phone was so SLOW.

HTC One S: software updates discontinued after only 6 months, but I kept it going for 4 years with custom ROMs. All custom ROMs, by the way, will always have an issue or two, because these are things you can’t write software tests for – you have to actually test it in the real world. Problems I remember: Bluetooth headphones changing pitch very slightly because the wifi antenna is retuning itself for a better reception (they use the same antenna after all), wifi not reconnecting when coming back into range (custom ROM issue), random reboots (custom ROM issue), no TRIM support on stock ROMs so you have to install a custom ROM if you don’t want your phone to slow down after a few months. On the plus side, the camera was everything I wanted and it was a sexy phone.

LG G5: tended to produce larger (3-6MB) photos even though pictures were blurred by the noise reduction algorithm. Then it got stolen.

Now I’ll dedicate a whole section to why iPhones are so amazing.

iOS: perfection

It took me a lot of fiddling with my HTC One S to understand myself – namely that even though I like fiddling with my computer, I don’t want to fiddle with my phone.

With the iPhone 5S, 6 and 6S, I got the whole package – a great camera; never felt slow; I actually used it and didn’t think about maintaining it.

camera such wow

Often times, you’re with somebody, they make a face or are in an interesting situation, and you want to take a picture. In the Android world you have to choose your phone carefully because all these could be bad:

Unlock the phone/swipe to camera shortcut: the camera app could start very slowly (500ms okay, 1s or more bad)

Camera takes a long time (1s or more) to focus.

Long delay after pressing the shutter button and the camera actually taking a picture (500ms or more)

Picture quality great in daylight (aren’t they all), terrible in certain conditions (still true today).

Apple may not be the absolute best in one dimension (focus speed, detail preservation, megapixel count, sensor etc) but it always strikes the best balance, consistently, since 2013. Android flagships often have great hardware but the software lets it down.

For instance, this picture was taken on a Redmi Note 9 Pro in the last minutes of an evening. The noise reduction algorithm in low light situations is too heavy handed and smudges the whole picture, especially the grass. Despite this loss in detail, the picture is a hefty 5MB.

Here’s another bad photo in a low light situation from the iPhone 6S. Notice how the carpet texture is preserved well – the Redmi would have reduced the carpet into a blurry mess. This photo is only 2MB.

other little details that just work

Did you know that digital compasses need to be calibrated? I don’t know how they do it, but on iPhones, it was always pointing in the right direction. The Redmi Note 9 led me down the wrong road several times, the compass being up to 90 degrees off. I calibrate it, and within a few minutes it’s wrong again. It’s such a small detail that ends up being a huge annoyance when traveling.

Redmi Note 9 Pro’s wifi/Bluetooth antenna is not the best. For instance, when paired with a cheap Bose QC25 bluetooth adapter, it can lose signal when I put my phone on the opposite side of my body!

iOS: after the honeymoon

After several years on iOS, I decided to experiment with Android again with a cheap phone, the Redmi Note 9 Pro, and I realized just what I had sacrificed.

evil capitalism

Apple is a company who clearly does NOT have your best interests in mind. 256GB upgrade to 512GB ($10, 20 difference at most)? $100, and no microSD card slot, and no audio jack either, because ‘courage’. With the iPhone 12 and 13, Apple even has software locks preventing you from swapping the battery or other components, and is pretending to go along with Right to Repair, while of course making it difficult to actually use its repair program.

Apple has an attitude problem.

me

It’s the attitude that’s the problem, when the company forces you to do things its way because you have no choice. In real life, we call them bullies.

Nobody else has your interests in mind either, except for the free open source software community. Even Microsoft places ads in Windows these days. Linux is the one place on the PC where I am not funneled towards someone’s agenda, thank god I learned to use it when I was a teenager.

Instagram does not have your best interests at heart. The official app is optimized to make you scroll forever and waste huge amounts of time. Thankfully, there is an alternative app on Android called Barinsta, which displays the same content in a normal way that doesn’t encourage you to waste an entire afternoon and doesn’t have ads. Such an app would never be allowed on the App Store, nor the Play Store.

Youtube does not have your best interests at heart. Just like Instagram, it encourages you to waste time with its Related Videos; autoplay is on by default; ads, popups that you can click on in the video area. I am not an idiot, I know that the difference between streaming a video and letting a video be downloaded so it can be played later offline is 0 – so why are they making me pay for this functionality with Youtube Premium? Thankfully there is Newpipe and Youtube Vanced (and its forks).

I could go on forever, but to make it short, imagine what the computing world would look like if software were written just for our (consumers) own benefit, and not just so companies could make money from us. That’s the free open source software movement, and F-Droid is full of apps which

1. don’t try to tack on a cloud/account/subscription model onto something which obviously doesn’t need one
2. isn’t some freemium hook and bait deal
3. can share files without relying only on cloud companies like Dropbox/OneDrive

Android as a smaller computer

I found myself using Android like a mini PC when I was too lazy to pull out my laptop.

Because it has a real filesystem, it’s easy to copy files (like movies or music) onto an Android phone and watch it on the plane. The iPhone would force me to use an app, which would force me to pay to download to watch/listen offline. Then I’d realize that I have no space left so I’ll have to give up on the idea.

iPhone vs Android – how to choose

Why would you buy an iPhone?

You go iPhone because you want things to just work – and you’ll make the necessary sacrifices to appease Apple so you can go about your daily life without having to think about your phone.

iPhone 12 costs $373 to produce, and sells for $800.

iPhone 12 Pro costs $406 to produce, and sells for $1000.

My iPhone 6S cost $200 in 2019, and I still got iOS 15 this year, so it’s definitely been worth it. At $1000, it’s probably still worth it, just because the software support is incredible – the 6S came out in 2015!

Still, I would never buy an iPhone new from Apple out of principle.

Why would you buy an Android?

You go Android because you want freedom and you don’t want to be shepherded into the behaviour patterns that big companies want you to do.

Expect little problems that would be a non-issue on iPhones, because nobody is as smart and detail oriented as Apple. Wait, what?

Think about it. Buying an Android flagship means you are paying $1000 for only 4 years of software updates (maximum, depends on vendor), and you’re losing the audio jack and microSD slot too! I might as well buy an iPhone already!

With dumbasses like these designing Android phones, and the fact that smartphone vendors don’t produce the processor nor Android, what do you expect?

Example: Xiaomi makes the Mi 12 Pro, which sells for $1100. I have the Redmi Note 9 Pro, which cost $180. The Gallery app has a problem: when I delete photos from the phone, the app still thinks the photos are there and there is no way to tell the app to refresh its database, short of reinstalling it!

In short, iPhones offer more than Androids at the $1000 price point. In the budget range, Androids are better value for money.

Removing the camera from the problem

I said before that no one is looking out for your interests. That includes smartphone review sites, who always make incredible claims about how much better photo quality is on this year’s crop of phones.

In daylight, quality is not a problem and hasn’t been since the iPhone 5S. The differentiator is low light performance. Smartphone cameras take multiple exposures in Night Mode to achieve better dynamic range. Noise reduction algorithm tuning is still an issue, because smaller sensors naturally do worse in the dark.

Below is a photo from the RX100 in a similarly dark, if not darker, situation. Imagine if the same amount of noise reduction/blurring from the P40 Pro were applied to this photo – would you still be able to read the GANT label on his jacket? How would his scarf’s texture look like then?

Let’s just cut the bullshit out and say: no smartphone camera is going to beat a bigger sensor, especially if things are moving at night. The devil is in the details, and what I’m realizing more these days is that the details are often left out by people with a vested interest.

The solution is obviously to always bring a dedicated compact camera with you. Except that of course, I rarely do, because the RX100 IV is still bulky, and it takes time to turn on.

Also you must consider the question: why do you take photos? It’s satisfying to take a good picture, but showing it to others gives me more happiness, even if it isn’t perfectly composed. It’s complicated with an external camera. You can’t share instantly, you have to copy it over first. It’s a different pace, slower, more deliberate.

Thanks to the way the world works, I’ll never get a mixture of an iPhone, Android, and a good camera.

It’s just easier to find a way to carry all 3 at once – I’ll figure it out one day…

Some time ago, in addition to being the trivia king of Fujitsu Primergy servers, I was also a master at finding used Macs for cheap on eBay. Nobody wants a Core 2 era Mac anymore, so naturally I bought a 2007 iMac 7,1 for 100EUR.

Now I have 8 Macs that I never use.

It was 100EUR because it would turn off randomly (power supply problem). Other than that it was perfect.

Life is so fast paced, nobody has time to repair broken things. I called a Mac repair store, even they didn’t have the time to repair it – too expensive to troubleshoot and take apart, they say, better to just buy another power supply. But they did leave me with a good tip – capacitors are the only thing that age and break down in a power supply.

It took me a year to finally get around to calling that repair store. I was constantly chasing the next nugget of knowledge, the next new hobby, the next acquisition. I have 14 jackets by the way, I just counted. I wanted to experience this, experience that.

When I bought that iMac, I was also chasing the next experience. But deep down, I was just chasing what I really enjoyed about computers when I was young – immersing myself in a project with no expectations whatsoever.

Pay attention to your self-talk (this goes for everything). “I’ve got to get it fixed by x date”, or “I’ve got to do this as quickly and efficiently as possible” sucks the fun out of it. Or “just gotta finish it quickly so I can move on to the next project” just misses the whole point of taking on such projects in the first place.

When I was ‘busy’ and ‘had to get all this done’, I didn’t get that much done and I was stressed and somewhat guilty at the end of the day.

When I told myself ‘you don’t have to do anything’, paradoxically I got so much done – because by not putting pressure on myself, I was able to reach flow.

And so it happens that after putting this off for a year, I suddenly have the time and energy to use an ESR meter on every capacitor in the power supply. Here’s the data, just to show the condition of the capacitors after 15 years:

C210 0.05O
C209 0.01O
C205 0.09O
C206 0.08O
C207 0.04O
C214 0.17O
C215 0.04O
C051 0.26O
C052 0.25O
C30? 0.19O 22uF 50V (obscured by goop)
C??? 0.70O 10uF 50V
C343 0.75O 10uF 50V
C204 0.03O
C406 47.0O! 10uF 100V 11mm -> replace with 10uF 160V 12.5mm tall Panasonic
C407 0.69O (out of circuit) 0.97O 22uF 50V
C213 0.06O
C212 0.07O
C211 0.10O
C004 0.26O
C057 0.20O

C406 and C407 were two small capacitors that lived under a large heatsink. C406 was totally gone, even though it wasn’t bulging. C407 had a in-circuit ESR of 0.97 Ohm. Compare this to C30?, which has the same capacitance, but is far away from the heatsink so it’s healthier at 0.19 Ohm. Although C407 is still “good”, it clearly has aged more.

Buying the ESR meter, desoldering pump, desoldering the TO220 component and the heatsink to get to the capacitors, finding an electronic parts shop willing to sell me only 1 of each capacitor… imagine if I had approached all of this with the mindset of “I need to get this done”.

I can’t wait for the capacitors to arrive.